Download 1. Methods

Towards Using Literature-based Discovery
to Explain Drug Adverse Effects
Dimitar HRISTOVSKIa,1, Anita BURGUN-PARENTHOINEb, Paul AVILLACHb and
Thomas C. RINDFLESCHc
a
Institute for Biostatistics and Medical Informatics, Medical faculty, University of
Ljubljana, Slovenia
b
University of Rennes, France
c
National Library of Medicine, NIH, Bethesda, USA
Abstract. We report on our preliminary research in using literature-based
discovery (LBD) to provide pharmacogenomic explanations for reported drug
adverse effects. The goal of LBD is to generate novel and potentially useful
hypotheses by analyzing the scientific literature and optionally some additional
resources. Our assumption is that drugs have effects on some genes and that these
genes are associated with the observed adverse effects. Therefore, by using LBD
we try to find genes that link the drugs with the reported adverse effects. These
genes can be used to provide insight into the processes causing the adverse effects.
Initial results show that our method has the potential to assist in explaining
reported drug adverse effects.
Keywords. Literature-based discovery, text mining, pharmacovigilance, adverse
drug effects, pharmacogenomics
Introduction
Pharmacovigilance is the discipline involved with the detection, understanding and
prevention of drug adverse effects, which are reported in specialized databases and in
the biomedical literature. The goal of our research is to find pharmacogenomic
explanations for known drug-adverse effect pairs. We try to find genes that link the
drugs with the adverse (side) effects. Our basic assumption is that the drugs have some
effect on some genes and that these genes are associated with the observed adverse
effects .
1. Methods
We use Literature-based Discovery (LBD) [1] to find explanations for (drug, adverse
effect) pairs. The goal of LBD is to generate novel hypotheses by analyzing the
literature and optionally other knowledge sources. LBD uses either of two basic
approaches: open discovery and closed discovery; both are based on a paradigm of
three related concepts: X, Y, and Z. In open discovery only the starting concept is
1
Corresponding Author. Dimitar Hristovski, PhD; Institute for Biostatistics and Medical Informatics,
Medical faculty, University of Ljubljana, Slovenia; Email: [email protected]
known. For example, if we want to find a new treatment for a given disease (X), we
first try to find (patho)physiological characteristics (Y) of the disease and then seek
drugs (Z) that can deal with these characteristics. In closed discovery both the starting
concept (X) and the end concept (Z) are known and we want to find intermediate,
linking concepts (Y). In any case, LBD is meant as a discovery support paradigm. LBD
generates hypotheses, but a knowledgeable human expert is needed for the
interpretation of these hypotheses. Our methodology is meant to assist an experienced
pharmacovigilance expert.
In our situation closed discovery is better suited because we work with known
adverse effects. In other words, the starting concept (Drug_X) is known as well as the
end concept (Adverse_effect_Z) and we want to find Genes_Y that somehow link the
drug with the adverse effects. By finding the linking genes, we provide explanation of
the statistically found association.
For our preliminary research we used a LBD tool called SemBT [2,3] available at
[4]. SemBT uses semantic relations extracted with the SemRep [5] natural language
processing system.
2. Results
To illustrate our methodology, we show here how we tried to find explanation for
recently published drug adverse effect, namely that lithium can trigger Brugada
syndrome [6, 7].
We have set lithium as the start concept X and Brugada syndrome as the target
concept Z. Then, with the SemBT tool we searched for genes or proteins that act as link
concepts between lithium and Brugada syndrome. We found three such genes or
proteins: Sodium Channel, Insulin and INS.
Table 1 shows aligned semantic relations extracted from the literature between
Lithium and Sodium Channel on the left side, and between Sodium Channel and
Brugada syndrome on the right side. The relations on the left side of Table 1 clearly
show that Lithium inhibits or blocks Sodium channels. And the right side shows that
Sodium Channel, or the gene encoding it, is etiologically related to Brugada syndrome.
Therefore, Sodium Channel acts as a link and provides an explanation for the reported
adverse drug effect that Lithium is associated with Brugada syndrome.
Table 1. Providing an explanation for the reported drug adverse effect (Lithium, Brugada syndrome) through
the linking concept Sodium Channel.
Aligned relations for Sodium Channel:
X-Relation-Y
Lithium
INHIBITS
Sodium
Channel
Lithium can unmask Brugada syndrome
through its ability to block sodium channels ,
even at subtherapeutic concentrations. (PMID:
20016437)
CONCLUSIONS: The widely used drug lithium
is a potent blocker of cardiac sodium channels
and may unmask patients with the Brugada
syndrome. (PMID: 16144991)
Y-Relation-Z
Sodium
Brugada
ASSOCIATED_WITH
Channel
syndrome
SCN5A, the gene encoding the alpha subunit of the
sodium channel , is the only gene thus far linked to
Brugada syndrome …(PMID: 16415541)
Mutations in SCN5A, a cardiac sodium channel gene,
have been recently associated with Brugada
syndrome . (PMID: 11960580)
Sodium
CAUSES
Brugada
Aligned relations for Sodium Channel:
X-Relation-Y
Because lithium is a potent blocker of cardiac
sodium channels , and given the critical
importance of sodium channels in pacemaker
activity, lithium-induced sodium channel
blockade is likely an important mechanism in
sinus node dysfunction. (PMID: 17347696)
Y-Relation-Z
Channel
syndrome
Loss of function mutations in SCN5A, encoding the
cardiac sodium channel , are one cause of the
Brugada syndrome ... (PMID: 16415376)
Changes in the sodium channel are responsible for
long QT syndrome, Brugada syndrome and
conduction defects. (PMID: 17497250)
Sodium
Channel
PREDISPOSES
Brugada
syndrome
A mutation in the human cardiac sodium channel
(E161K) contributes to sick sinus syndrome,
conduction disease and Brugada syndrome in two
families. (PMID: 15910881)
3. Discussion
In order for our methodology to work, a few preconditions need to be met, especially
for exploiting it on a large scale, i.e. providing a generally available tool on the web.
One is concept mapping. Specialized adverse effects reporting databases, such as
AERS, mostly use MEDDRA, and we use UMLS concepts. When they match directly
there is no problem, but if they do not, we have to map MEDDRA concepts to UMLS.
We can work only with drugs and adverse effects for which enough publications have
been written. We need enough publications because all the semantic relations are
extracted from Medline citations (the title and abstract ONLY). Later we plan to extract
semantic relations with SemRep from the full text of the articles.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Swanson DR. Fish oil, Raynaud’s syndrome, and undiscovered public knowledge. Perspect BiolMed
1986;30(1):7–18.
Hristovski D, Kastrin A, Peterlin B, Rindflesch TC. Semantic relations for interpreting DNA
microarray data. AMIA Annu Symp Proc. 2009 Nov 14;2009:255-9.
Hristovski D, Kastrin A, Peterlin B, Rindflesch TC. Combining Semantic Relations and DNA
Microarray Data for Novel Hypotheses Generation. In: BLASCHKE, Christian (ed), SHATKAY, Hagit
(ed). Linking literature, information, and knowledge for biology. (Lecture notes in bioinformatics), pp.
53-61. Springer: Heidelberg, 2010.
http://sembt.mf.uni-lj.si (last accessed 01.03.2012)
Rindflesch TC, Fiszman M. The interaction of domain knowledge and linguistic structure in natural
language processing: interpreting hypernymic propositions in biomedical text. J Biomed Inform. 2003
Dec;36(6):462-77.
http://www.fda.gov/Drugs/GuidanceComplianceRegulatoryInformation/Surveillance/AdverseDrugEffe
cts/ucm237585.htm (last accessed 01.03.2012)
http://www.medscape.com/viewarticle/712193_6 (last accessed 01.03.2012)